Looks as though submitting tar.gz files results in a single partition, at least early on in spark workflow. By contrast, pointing to a directory with XML files results in the number of partititions as there are files (e.g. 1984 in one case).
It might make sense to rationalize this a bit and partition after reading files: partition up from archive files, slim down from large number of files.
Looks as though submitting
tar.gz
files results in a single partition, at least early on in spark workflow. By contrast, pointing to a directory with XML files results in the number of partititions as there are files (e.g.1984
in one case).It might make sense to rationalize this a bit and partition after reading files: partition up from archive files, slim down from large number of files.